ok, i found some more diagnostic messages in /sys/log/sshdebug:

        p9 Jan 21 10:55:48 netssh: client user <nil>@192.168.1.12 id 0 id 
string `SSH-2.0-OpenSSH_6.7p1-hpn14v5
        p9 Jan 21 10:55:48 netssh: client user <nil>@192.168.1.12 id 0 sent KEX 
algs: diffie-hellman-group1-sha1,diffie-hellman-group14-sha1
        …
        p9 Jan 21 10:55:49 netssh: client user <nil>@192.168.1.12 id 0 using 
diffie-hellman-group14-sha1 Kex algorithm and ssh-rsa PKA

in contrast to:
        p9 Jan 21 10:57:31 netssh: client user <nil>@192.168.122.6 id 0 id 
string `SSH-2.0-OpenSSH_6.6.1p1-hpn14v5      
        …
        p9 Jan 21 10:57:31 netssh: client user <nil>@192.168.122.6 id 0 using 
diffie-hellman-group1-sha1 Kex algorithm and ssh-rsa PKA

The problem might be that `dh.c` has an empty implementation of `dh_client142`

        Kex dh1sha1 = {
                "diffie-hellman-group1-sha1",
                dh_server1,
                dh_client11,
                dh_client12
        };
        
        Kex dh14sha1 = {
                "diffie-hellman-group14-sha1",
                dh_server14,
                dh_client141,
                dh_client142
        };


> Hi,
> 
> the netssh key exchange seems to be incompatible with openssh-6.7. 
> 
> I installed a new version of openssh on a gentoo host recently, that 
> automatically came in as a stable update package for a gentoo-amd64 system:
> 
>       OpenSSH_6.7p1-hpn14v5, OpenSSL 1.0.1k 8 Jan 2015
> 
> When calling this system with a plan9 (legacy) ssh2, the netssh process does 
> not provide any data in /net/ssh/keys. The read at 
> /sys/src/cmd/ssh2/ssh2.c:/^keyproc/+19, reads n=0 bytes when connecting to 
> the version of OpenSSH above.
> 
> I don't understand enough of the netssh keyfile infrastructure to debug this 
> logistic behaviour of /net/ssh/keys.
> 
> A downgrade to
> 
>       OpenSSH_6.6p1-hpn14v4, OpenSSL 1.0.1k 8 Jan 2015
> 
> gives me ssh access to the gentoo system again.
> 
> If I find out more, I will post a followup. But maybe it would be helpfull if 
> someone with more insight into netssh tries to resolve this bug.
> 
> regards,
> 
> ingo krabbe



Reply via email to